We present a new open-source parallel corpus consisting of news articles collected from the Bianet magazine, an online newspaper that publishes Turkish news, often along with their translations in English and Kurdish. In this paper, we describe the collection process of the corpus and its statistical properties. We validate the benefit of using the Bianet corpus by evaluating bilingual and multilingual neural machine translation models in English-Turkish and English-Kurdish directions.